Beruflich Dokumente
Kultur Dokumente
FINAL REPORT
- October 2002 -
This report was prepared in cooperation with the State of Florida Department of Transportation and the U.S.
Department of Transportation. The report does not constitute a design standard, specification, or regulation.
The opinions, findings, and conclusions expressed in this publication are those of the authors in the course and
scope of employment by the University of Central Florida and not necessarily those of the Florida Department
of Transportation or the U.S. Department of Transportation.
ACKNOWLEDGEMENT
This project was funded in part by the Florida Department of Transportation (Contract BC421) and the Department of Civil and Environmental Engineering at the University of
Central Florida. Special thanks are due to former UCF graduate student Bibo Zhang, who
worked as a research assistant on this project.
ii
TABLE OF CONTENTS
1
Introduction....................................................................................................................... 1
1.1
Background ............................................................................................................... 1
1.2
1.3
Report Organization.................................................................................................. 4
Introduction............................................................................................................... 5
2.2
2.3
2.4
2.5
Elements Library..................................................................................................... 10
2.5.1
2.5.2
2.6
2.6.1
2.6.2
2.6.3
2.6.4
2.7
2.7.1
iii
2.7.2
2.7.3
Background ............................................................................................................. 38
3.2
3.3
3.3.1
3.3.2
3.3.3
3.3.4
3.4
3.5
Results..................................................................................................................... 52
3.5.1
3.5.2
3.5.3
Effect of ...................................................................................................... 52
3.6
3.7
Introduction............................................................................................................. 59
4.2
iv
4.3
4.3.1
4.3.2
4.4
Parametric Study..................................................................................................... 66
4.5
Introduction............................................................................................................. 71
5.2
5.2.1
Strength ........................................................................................................... 71
5.2.2
Feasibility........................................................................................................ 72
5.2.3
Accessibility.................................................................................................... 72
5.2.4
5.2.5
5.2.6
5.3
5.4
5.4.1
5.4.2
5.5
5.6
REFERENCES ............................................................................................................... 84
9.2
vi
LIST OF FIGURES
Figure 1.1: View showing diaphragms connecting two box girders in an interchange............ 2
Figure 1.2: Interior of a box girder. Note the vertical cross-frames (comprised of sloping truss
members), web stiffeners, bottom flange stiffeners, and top bracing............................... 3
Figure 2.1: Flow chart of computer program (part I)................................................................ 7
Figure 2.2: Flow chart of computer program (part II) .............................................................. 8
Figure 2.3: Plan view of the grid model ................................................................................... 9
Figure 2.4: Setting 3-D viewing parameters ............................................................................. 9
Figure 2.5: Isometric 3-D view of deformed shape ................................................................ 10
Figure 2.6: Local coordinate system for 6-DOF element ....................................................... 11
Figure 2.7: Local coordinate system for 7-DOF element ....................................................... 15
Figure 2.8: Warping of a closed thin-walled cross-section..................................................... 15
Figure 2.9: A general element showing non-coincident shear center and centroid ................ 23
Figure 2.10: Local and global CS for handling of restrained DOFs....................................... 25
Figure 2.11: Normal stress distribution under warping for a I-section beam ......................... 25
Figure 2.12: Bimoment distribution along the entire length of Bridge 521 (program results)30
Figure 2.13: Bimoment distribution along the entire length of Bridge 521 (ABAQUS results)
......................................................................................................................................... 30
Figure 2.14: Dimensions of the cross section for verification example (in inches) ............... 31
Figure 2.15: Loading of verification example ........................................................................ 31
Figure 2.16: 3D-view of Model A in SAP 2000..................................................................... 32
vii
viii
Figure 4.4: Dimensions for calculation of total torsional constant based on BEF ................. 63
Figure 4.5: Verification results for proposed grillage model.................................................. 65
Figure 5.1: Alternative access hole locations ......................................................................... 72
Figure 5.2: Low stress regions for idealized bridge (a - Approach I, b - Approach II). ......... 76
Figure 5.3: (a) Distribution of fatigue stresses. Empty regions are areas in compression and
therefore not affected by fatigue. (b) Regions not critical for fatigue considerations
(entire bridge).................................................................................................................. 77
Figure 5.4: Regions that satisfy both normal stress criteria (<33% regions in Figure 5.2-a and
fatigue stress criteria in Fig. 5.3-b). Shaded regions are suitable for access hole
placement. ....................................................................................................................... 78
Figure 5.5: General view of finite element mesh.................................................................... 80
Figure 5.6: Idealized stress-strain relationships for steel and concrete. ................................. 80
Figure 5.7: Bottom view of bridge model showing hole location in a minimally stressed
region. ............................................................................................................................. 81
Figure 8.1: A general closed thin-walled cross-section .......................................................... 89
Figure 8.2: diagrams of closed and open section parts ...................................................... 89
ix
LIST OF TABLES
Table 2.1: M w verification results Program vs. closed form solutions ............................... 29
Table 2.2: Summary of SAP2000 shell models ...................................................................... 33
Table 2.3: Maximum normal stress from program and Models A, B, and C ......................... 36
Table 3.1: Summary data for analyzed existing bridges......................................................... 40
Table 3.2: Definitions of Warping Stress Ratio ( WSR ) ......................................................... 51
Table 4.1: Details of variables considered in parametric study .............................................. 67
Table 4.2: Comparison of distribution factor results (L=25 m).............................................. 68
Table 4.3: Comparison of distribution factor results (L=50 m).............................................. 69
Table 4.4: Comparison of distribution factor results (L=100 m)............................................ 70
Table 5.1: Summary of advantages and disadvantages of access hole alternatives................ 74
EXECUTIVE SUMMARY
Problem Statement
Composite steel-concrete box girders are commonly used in curved bridges, interchanges,
and ramps. Curved composite box girders have a number of unique qualities that make them
suitable for such applications including: 1) their structural efficiency allows designers to
build long slender bridges that have an aesthetically pleasing appearance, and 2) composite
box girders are particularly strong in torsion and efficiently resist the large torsional demands
created by horizontal bridge curvature and vehicle centrifugal forces.
Analysis and design of curved composite box bridges is complicated by many factors
including: composite interaction between the concrete deck and steel U-girder, local buckling
of the thin steel walls making up the box, torsional warping, distortional warping, interaction
between different kinds of cross-sectional forces, and the effect of horizontal bridge
curvature on both local and global behavior.
Current codes pertaining to analysis and design of curved composite girders are mostly based
on experimental and analytical research conducted over 30 years ago as part of project CURT
(Consortium of University Research Teams) funded by the Federal Highway Administration
(FHWA). A new Curved Steel Bridge Research (CSBR) project is currently being conducted
under the auspices of the FHWA. Although the CSBR project is expected to provide much
needed information on behavior, analysis and design of curved composite bridges, it focuses
more on I-girders than on box girders.
Objectives
The overall objective of the research reported herein is to provide information that
complements existing data and that will be useful for formulating comprehensive design
guidelines for composite curved box girders. Specific objectives include:
Investigate and quantify the effect of nonuniform torsion on the behavior and design of
existing curved box girders.
Study the adequacy of existing distribution factors for curved box girder bridges.
xi
Summary of Work
A computer program for simulating the behavior of curved box girders is developed. The
program is graphically interactive and features a general purpose beam-column element that
can account for the effect of warping. The developed program is used to conduct a detailed
investigation of warping related stresses in eighteen existing box girder bridges chosen from
the Florida Department of Transportation inventory. The bridges are carefully selected to
cover a wide range of design parameters including horizontal curvature, cross sectional
properties, and number of spans. They were designed by different firms and were constructed
at different times and are considered to be representative of current design practice. Forces
are evaluated from analyses that account for the construction sequence and the effect of
warping. Loading is considered following the 1998 AASHTO-LRFD provisions. The
differences between stresses obtained taking into account warping and those calculated by
ignoring warping are used to evaluate the effect of warping.
Another study was undertaken to investigate load distribution factors promoted by current
specifications. Single girder and detailed grillage models were created for a variety of bridges
and analyzed suing the developed program. The parameters investigated are the number of
girders, roadway width expressed by number of lanes, girder spacing, span length, and radius
of horizontal curvature. The distribution factor results were compared with those obtained
using the equations recommended by AASTHO in the commentary of the guide specification
for horizontally curved bridges. Results show that the recommended equation overestimates
the distribution factor by as much as 25% with an average of about 15%. In some cases,
AASHTOs equation yielded unconservative results.
Access hatches (holes) in curved box girder bridges are usually provided in the bottom flange
immediately before or after an expansion joint. If additional access hatches are required after
the bridge is built, they must be placed in such a way that 1) they satisfy important practical
constraints such as feasibility, accessibility, water leakage, traffic impact, and unauthorized
access; 2) they do not adversely affect the structural behavior of the bridge, i.e. their
installation should not impair serviceability nor decrease ultimate strength or fatigue life.
Both issues are discussed and approaches that are suitable for identifying appropriate
locations for access hole placement are identified.
xii
are complicated and time consuming. Additional work is needed to define relevant
parameters that can be used to identify bridges where warping calculations are warranted.
The authors also believe that there is a need for a validated approximate design method that
accounts for the effect of warping, without which it is hard to envision designers performing
detailed analyses such as those presented here.
Existing distribution factor equations are in need of substantial improvement. More
variables should be considered in developing the new expressions including torsional rigidity
of the box.
Access hatches can be installed without additional strengthening in low stress regions in
the bottom steel flange. Low stress regions can be found using the program developed in this
research.
xiii
1 INTRODUCTION
1.1 Background
Composite steel-concrete box girders are commonly used in curved bridges, interchanges,
and ramps (Fig. 1.1). Curved composite box girders have a number of unique qualities that
make them suitable for such applications including: 1) their structural efficiency allows
designers to build long slender bridges that have an aesthetically pleasing appearance, and 2)
composite box girders are particularly strong in torsion and can be easily designed to resist
the large torsional demands created by horizontal bridge curvature and vehicle centrifugal
forces.
Curved composite box girder bridges are generally comprised of one or more steel U-girders
attached to a concrete deck through shear connectors. Diaphragms connect individual steel
U-girders periodically along the length to ensure that the bridge system behaves as a unit
(Fig. 1.1). The cross-section of a steel box is flexible (i.e. can distort) in the cross-wise
direction and must be stiffened with cross-frames that are installed in between the
diaphragms to prevent distortion (Fig. 1.2). Web and bottom plate stiffeners are required to
improve stability of the relatively thin steel plates that make up the steel box. During
construction, overall stability and torsional rigidity of the girder are enhanced by using top
bracing members. These bracing members become unimportant once the concrete decks
hardens, but are usually left in place anyway (Fig. 1.2).
Analysis and design of curved composite box bridges is complicated by many factors
including: composite interaction between the concrete deck and steel U-girder, local buckling
of the thin steel walls making up the box, torsional warping, distortional warping, interaction
between different kinds of cross-sectional forces, and the effect of horizontal bridge
curvature on both local and global behavior. Existing literature contains extensive
information about the analysis, behavior, and design of horizontally curved composite box
girder bridges. General theories can be found in textbooks (e.g. Guohao 1987 and Nakai and
Yoo 1988) and a comprehensive survey of experimental and analytical work on curved steel
girders (including box girders) can be found in Zureick et al (1994) and Sennah and Kennedy
(2001 and 2002).
Current codes pertaining to analysis and design of curved composite girders include
AASHTO's LRFD Bridge Design Specifications (1998) and Guide Specifications for
Horizontally Curved Highway Bridges (1997). Provisions in these specifications are mostly
based on experimental and analytical research conducted over 30 years ago as part of project
CURT (Consortium of University Research Teams) funded by the Federal Highway
Administration (FHWA). A new Curved Steel Bridge Research (CSBR) project is currently
1
being conducted under the auspices of the FHWA with the following objectives (Zureik et al
2000): 1) gain a better understanding of the behavior of curved steel girders through large
scale tests and numerical modeling, and 2) update existing design provisions. Although the
CSBR project is expected to provide much needed information on behavior, analysis and
design of curved composite bridges, it focuses more on I-girders than on box girders.
Intermediate
Diaphragms
Figure 1.1: View showing diaphragms connecting two box girders in an interchange
Top Bracing
Stiffeners
Cross Frame
Figure 1.2: Interior of a box girder. Note the vertical cross-frames (comprised of
sloping truss members), web stiffeners, bottom flange stiffeners, and top bracing.
as well as existing structures to improve accessibility. The results of a study that addresses
this issue are summarized in Chapter 5.
Two other topics of significance to the design of curved composite box girder bridges are
also discussed in this report. First, the issue of warping (non-uniform torsion) is investigated
in a detailed and comprehensive manner. Warping has been rarely the focus of researchers in
the recent literature. Code provisions related to warping are based on old studies, which to
say the least, offer little guidance to designers. Second, live load (LL) distribution factors are
investigated with the help of a computer program specifically developed for this purpose.
AASHTO-LRFD (1998) has adopted the LL distribution factor approach in its provisions to
simplify the design process. This approach, while convenient for designers, has its obvious
limitations. Since bridges are categorized into groups based on structural systems, each group
is subject to a set of formulas that provides the LL distribution factor. Within each group,
several parameters are deemed important and the code attempts to calibrate equations that are
applicable for a wide range of values for these parameters. While AASHTO-LRFD (1998)
intends for the distribution factors to be conservative, it is inevitable that in some cases the
suggested expressions will yield unconservative estimates. Several studies have been recently
published on the distribution factors for girder bridges, attempting to enhance the available
expressions. However, very little has been published on distribution factors for composite
box girder bridges.
2.1 Introduction
Early on in the project, a decision had to be made about what was the best analysis tool for
handling the tasks at hand. The two options available were: 1) to use a commercially
available package, or 2) to develop a special program tailored to the needs of the project.
Commercial packages may be categorized into two main groups. The first group is researchoriented packages that often provide an abundant array of special feature including providing
elements with warping capabilities. These packages are not geared towards design in
general, and bridge design in particular. For example, they do not offer a loading module
according to the latest code provisions. Using such packages would entail an enormous
amount of work to provide the details of the bridge loading and special details of connections
between different elements. Packages geared towards design are basically finite element
implementations with strict code interpretations. The finite element modules are usually
simpler than those in research packages. In other words, they follow the latest code
provisions, but lack the ability to incorporate special effects such as warping. Some
packages with warping capabilities implement a simplified approximate method in their
analyses.
Based on the limitations stated earlier, a decision was made to develop a computer program
for the analysis of curved composite box girder bridges. While this choice involves more
work than resorting to available commercial packages, it was deemed necessary because of
the nature of the studies involved.
This chapter describes the computer program that was developed throughout the course of
this project. It is a finite element program that is specifically tailored for the analysis of
curved composite box girder bridges. It is developed in the VisualBASIC environment
which allowed providing it with a graphical user interface (GUI) to facilitate its use. Several
issues were encountered while developing the program, all of which have been solved and
are also described in this chapter.
be straight or curved. The focus in this type of analysis is on forces rather than on stresses.
The second category of analysis techniques is the micro model approach, examples of which
are the finite element method or finite strip method. In these methods, the bridge system is
discretized into a number of continuum elements, and the emphasis is on stresses and
corresponding strains.
Micro models are more rational than macro models, and provide more detailed information.
However, they are difficult to setup and analyze, and are usually used for research purposes
or to validate new designs. Macro-modeling techniques on the other hand, are simple to
implement and since they yield reasonable results, are commonly used by practitioners and
researchers. In fact, the plane grid method was used to calibrate the load distribution factors
in the current AASHTO-LRFD specifications (1998). The analysis method adopted in the
program is the space frame approach and falls under the macro-model category.
truck using a magnification factor of 200. The model is viewed with a plan angle of 150o and
an elevation angle of 15o . This view is also useful for checking the model for input errors.
start
x, y , I x , I y , I , S x , S y , S
Form
transformation matrices
T
Form element stiffness
matrix in global CS
global
k element
local
k element
global
Assemble k element
in
Global Stiffness Matrix
global
K structure
global
Adjust K structure
for
constraints and
boundary conditions
global
end
This element is the classical three-dimensional frame element with 6 DOFs per node. The
T
DOFs are {u v w x y z } in the local CS of the element as can be seen in Fig. 2.6.
The element stiffness matrix has an order of 12x12.
10
y2
y2
x2
z2
'
x2
z2
x
z
y1
x1
x1
y1
z1
z1
Stiffness relations for this element can be derived using the direct method or the virtual work
principle. In the direct method, each DOF is released while other DOFs are restrained; one at
a time. A unit deformation is then imposed on the released DOF resulting in reactions at the
restrained DOFs. The reactions constitute the stiffness terms of the column elements
corresponding to the released DOF. This straight forward method is suitable for simple
elements where the reactions due to the imposed unit deformation can be easily quantified.
The virtual work principle is based on the concept of energy conservation. It is a well known
approach, but its essential features will be outlined here for completeness A virtual
displacement, , is imposed on the structure in this case the element. The external work
done to impose this virtual displacement has to be equal to the internal strain energy,
W = Wext Wint = 0
(2.1)
The virtual displacement field imposed on the element follows an assumed displacement
field characterized by shape functions, [N ]. The relationship between the displacement, , at
any point and nodal displacements for a general element with n DOFs is given by
n
= N 1 1 + N 2 2 + ... + N n n = N i i = N {}
(2.2)
i =1
where N i is the shape function corresponding to the i th DOF. The strain energy is calculated
with the help of the strain displacement matrix, [B ] , which contains derivatives of the shape
functions and relates generalized strains to nodal displacements. For example, the
11
generalized axial strain is obtained using the first derivative of [N axial ] , where [N axial ]
contains the shape functions pertaining to the axial degrees of freedom.
'
]{}
axial = [N axial
(2.3)
which establishes the fact that the strain-displacement matrix for axial deformation DOFs is
'
[Baxial ] = [N axial
]
(2.4)
A relationship involving the second derivative of [N ] yields [B ] matrix for flexural terms. In
general, the strain energy due to the imposed virtual displacement field is given as
(2.5)
vol
where [D ] is the constitutive matrix, which differs based on the DOF under considerations.
For example, [D ] for axial deformations is simply the modulus of elasticity. The external
work done to impose the virtual displacement field, {} , is given by
Wext = {F }
(2.6)
vol
[B ]T [D ][B ]d (vol){} = {F }
(2.7)
Comparing Eq. 2.7 to the general stiffness relationship provides the following expression for
the stiffness matrix.
vol
(2.8)
The accuracy of the derived stiffness matrix depends on the quality of the assumed shape
functions representing various displacement fields. Axial deformations are often represented
by linear shape functions satisfying preliminary strength of materials requirements. The axial
deformation, u , at any point within the element is given by
12
u
u = N axial 1
u 2
(2.9)
where for an element of length L, the assumed shape functions pertaining to the axial degrees
of freedom are
N axial =
Lx
L
x
L
(2.10)
Linear shape functions are also assumed for torsion and the twisting angle DOF.
x = N Torsional x1
x 2
(2.11)
L x
L
(2.12)
N =
x
L
where [N Torsional ] has the shape functions pertaining to the torsional degrees of freedom.
Hermitian cubic polynomials are assumed for flexural related DOFs.
v = N1
N2
N3
v1
N 4 x1
v2
x 2
(2.13)
where
N1
N
N = 2
N3
N 4
2
3
x
x
1 3 + 2
L
L
2
3
x
x
3 2
L
L
=
2
x 1
L
2
x x x
L
L
(2.14)
13
An identical set of flexural shape functions is also assumed for the other (weak) axis of
bending. The previous derivation is totally uncoupled; i.e. torsion is not affected by bending
and axial DOFs are not affected by torsion, and so on. This implies that the derived stiffness
matrix is formulated for small deformations. Large deformations in which secondary effects
take place are not accounted for this research.
2.5.2 Seven DOF Implementation
One of the goals of this study is to investigate the warping behavior of curved composite box
girder bridges. This element is developed for that purpose. Warping is accounted for through
an added seventh DOF. The additional DOF is the first derivative of the twisting angle, x .
At each node, the DOFs for this element become {u v w x y z x } in the local
T
coordinate system (CS) of the element as can be seen in Fig. 2.7. Other implementations of
warping are possible; however, the formulation presented next captures the essential aspects
of the behavior. It is well known that the St. Venant torsion, Ts , which is often referred to as
uniform torsion, is normally expressed in terms of the torsion constant, J , and the shear
modulus, G , and the twisting angle, x , as follows
Ts =
GJ x
= GJ x
L
(2.15)
The bimoment, M , which accompanies torsion and causes out of plane deformations as
those shown in Fig. 2.8 is a function of the warping constant, I , the modulus of elasticity,
E , and the second derivative of the twisting angle, x
M = EI w x
(2.16)
The total torsion may now be viewed as comprised of two components; the uniform torsion,
Ts , and the non-uniform torsion, T . It can be proven that the non-uniform torsion is the first
derivative of the bimoment
T =
d
M = EI x
dx
(2.17)
14
y2
y2
x2
z2
'
x2
x2
z2
x
z
y1
x1
y1
z1
z1
x1
'
x1
MT
MT
M x = T = Ts + T = GJ x' EI w x'''
(2.18)
15
Formulating the element stiffness matrix equations by employing virtual work principles
leads to the following expressions for the internal and external virtual work
Wext = M x + T x
L
(2.19)
(2.20)
can now be
expressed as
[k ] =
(2.21)
If the displacement field for the twisting angle is assumed to be a cubic polynomial as
follows
x = a1 + a 2 x + a3 x 2 + a 4 x 3
(2.22)
x = a2 + 2a3 x + 3a4 x 2
(2.23)
The coefficients, ai , in the previous expressions can be found using the appropriate boundary
conditions, which leads to the same expressions as those obtained previously for bending
(Eqs. 2.13 and 2.14). Substituting for the coefficients in the previous equations yields the
following submatrix for torsion related DOFs.
6
6
5L 5L
5L
[k ] = GJ
sym
1
10
1
10
2L
15
1
12
12
10
2
2
1
L
L
EI
12
10 +
L
L2
L
30
sym
2L
15
16
6
L
6
L
4
6
L
6
L
2
4
(2.24)
EI
, both terms in Eq. 2.24 can be lumped into the
GJL2
following matrix
1
L + 6
5 + 12 5 + 12 L 10 + 6
10
6
1
1
+ 12 L + 6 L + 6
GJ
5
10
10
[k ] =
2
L
L2 + 4 L2
+ 2
15
30
L2 + 4
sym
15
(2.25)
This submatrix is added to the other uncoupled DOFs to obtain the full 14x14 stiffness
matrix. The assumption of uncoupled DOFs does not always hold and special treatments are
necessary. The following sections illustrate some of the special treatments implemented in
the program.
Shear deformations are quite small in most civil engineering applications and they are
therefore often ignored with no impact on the quality of results. However, shear deformations
may be dominant in some situations where bending moments are small compared to shear
forces acting on the member. This is normally true for short span beams. In a structural
system such as a curved box girder bridge, this is also true for diaphragms connecting box
girders in the transverse directions. To account for shear deformations, the concept of
equivalent shear area, AS , is used. According to this concept, the applied shear force, Fy , is
equal to the equivalent shear area multiplied by the shear stress at the centroid of the cross
section. The component of complementary internal virtual work due to shear can be
expressed as
l
F y 1 F y
Wint( S ) =
AS G AS
0
F F L
AS dx = y y
AS G
The total strain energy is equal to the summation of the bending and shear components
17
(2.26)
L3
L
+
3EI z AS G
M z
L2
2 EI z
L2
2 EI z Fy
L M z
EI z
(2.27)
The middle term of the above expression is the flexibility matrix taking into account shear
deformations. The stiffness matrix is then obtained by transforming the flexibility matrix
which leads to the following (McGuire et al. 2000)
F y1
EI z
z1
2
F
y 2 L L +
12
M y 2
symmetric
L
2
L2
L
+
3
2
1
v1
2
L
6
1
L v2
2
2
L2
+
3
L
2
(2.28)
= z =
EI z
AS G
(2.29)
Satisfactory results were obtained when AS was taken equal to the area of the webs in the
appropriate direction. The resulting stiffness matrix for the 6-DOF implementation is now
given as
K 0
K (0e ) = (0e )(1)
K ( e )(3)
K (0e )( 2)
K (0e )( 4)
(2.30)
18
K(0e)(1)
K(0e)(2)
EA
0
L
12EIz
0
3
L +12Lz
0
0
=
0
0
0
0
6EIz
0
2
L +12z
EA
0
L
12EIz
0
3
L +12Lz
0
0
=
0
0
0
0
6EIz
0
L2 +12z
12EI y
L3 +12L y
0
6EI y
L2 +12 y
0
0
6EIz
L +12z
2
6EI y
0
GJ
L
EI y L2 / 3 + y
L L2 / 12+ y
12EI y
L3 +12L y
0
6EI y
GJ
L
0
L2 +12 y
0
19
L2 +12 y
0
EIz L2 / 3 +z
L L2 / 12+z
6EI y
L2 +12 y
0
EI y L2 / 3 + y
L L2 / 12+ y
0
0
6EIz
L +12z
2
0
0
0
EIz L2 / 6 z
L L2 / 12+z
(2.31)
(2.32)
K(0e)(3)
K(0e)(4)
EA
0
L
12EIz
0
3
L +12Lz
0
0
=
0
0
0
0
6EIz
0
2
L +12z
EA
0
L
12EIz
0
3
L +12Lz
0
0
=
0
0
0
0
6EIz
0
L2 +12z
12EI y
0
6EI y
0
L L2 / 12+ y
0
L3 +12L y
0
6EI y
L2 +12 y
0
GJ
L
0
0
20
EI y L2 / 6 y
12EI y
L2 +12z
L2 +12 y
GJ
L
0
L2 +12 y
6EIz
6EI y
L3 +12L y
6EI y
L2 +12 y
0
EI y L2 / 3 + y
L L2 / 12+ y
0
0
EIz L2 / 6 z
L L2 / 12+z
0
6EIz
L2 +12z
0
0
0
EIz L2 / 3 +z
L L2 / 12+z
(2.33)
(2.34)
I y and I z
E and G
y =
EI y
ASz G
is the shear deformation term in the z direction of the local CS based on the
shear area ASz
z =
EI z
ASy G
is the shear deformation term in the z direction of the local CS based on the
shear area ASy
0
(e)
K (0e )(1)
= 0
K ( e )(3)
K (0e )( 2)
K (0e )( 4)
(2.35)
K(e)(1)
EA
0
L
12EI z
0
L3 + 12Lz
0
0
0
= 0
0
0
6EIz
0
2
L + 12z
0
0
6EIz
L2 + 12z
12EI y
L + 12L y
3
0
6EI y
L2 + 12 y
6EI y
L2 + 12 y
GJ 6
+ 12
L 5
0
EI y L2 / 3 + y
L L2 / 12 + y
0
0
0
GJ + 6
10
EI z L2 / 3 + z
L L2 / 12 + z
21
(2.36)
1
GJ + 6
10
GJL + 4
15
EA
0
L
12EIz
0
3
+ 12Lz
L
0
0
0
K(e)(2) = 0
0
0
6EIz
0
L2 + 12z
0
0
K(e)(3)
K(e)(4)
EA
0
L
12EIz
0
3
L +12Lz
0
0
0
= 0
0
0
6EIz
0
2
+12z
L
0
0
EA
0
L
12EIz
0
L3 +12Lz
0
0
0
= 0
0
0
6EIz
0
L2 +12z
0
0
6EIz
L2 + 12z
12EIy
L + 12Ly
0
6EIy
GJ 6
+ 12
L 5
6EIy
L2 + 12y
L2 + 12y
0
EIy L2 / 3 + y
L L2 / 12 + y
GJ + 6
10
EIz L2 / 6 z
L L2 / 12 + z
12EIy
L +12Ly
0
6EIy
L2 +12y
6EIy
L +12y
2
GJ 6
+12
L 5
0
EIy L2 / 6 y
L L2 / 12+y
0
6EIz
L2 +12z
0
0
0
GJ + 6
10
EIz L2 / 6 z
L L2 / 12+z
6EIz
L2 +12z
12EI y
L +12L y
3
0
6EI y
L2 +12 y
6EI y
L +12 y
2
GJ 6
+12
L 5
0
EI y L2 / 3 + y
L L2 / 12+ y
0
0
0
GJ + 6
10
EIz L2 / 3 +z
L L2 / 12+z
22
(2.37)
1
GJ + 6
10
GJL + 2
30
0
(2.38)
1
GJ + 6
10
GJL + 2
30
0
(2.39)
1
GJ + 6
10
GJL + 4
15
The previous derivation of the stiffness matrix is based on the assumption that the centroid
and the shear center of the cross section coincide. This assumption is true for many
engineering shapes, however, it is not the case for box girder cross sections. The eccentricity
between the centroid and shear center is relatively small for closed cross sections. But as
will be seen later, the cross section is not fully closed (quasi-closed) during construction. At
this stage, the shear center often falls below the bottom flange while the centroid is usually
within the plates (flanges and webs). Hence, a large eccentricity is created. Not accounting
for this eccentricity is inaccurate modeling. Therefore, the derived stiffness matrix is treated
by a special transformation for that purpose. For a general cross section with non-coincident
shear center and centroid as in Fig. 2.9
Figure 2.9: A general element showing non-coincident shear center and centroid
The eccentricity between the shear center and centroid leads to a coupling between torsion
and bending DOFs. It is assumed that the local coordinate system x passes through the
centroid line of the straight beam, and that the y and z represent the weak and strong bending
axes, respectively. A shear force acting at the centroid of the cross section, but not the shear
center, generates extra torsion because of the eccentricity. The effect of the eccentricity (0,
dy, dz) is accounted for using the following transformation
T
K SC = TSC * K * TSC
(2.40)
23
T SC
1
0
0
0
0
0
=
0
0
0
0
0
0
corresponding
to
0
0
0
0
0
0
0
0
0
0
0
0
1
(2.41)
following
order
0
1
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
dz
0
0
0
0
dy
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
dz
0
0
1
dy
0
0
0
1
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
the
14
degree
of
u1 , v1 , w1 , x1 , y1 , z1 , x' 1 , u 2 , v2 , w2 , x 2 , y 2 , z 2 , x' 2
freedom
T
in
the
Due to the nature of curved bridges, it was necessary to handle support DOFs in a special
manner. Supports of curved girders restrain movement in the vertical and horizontal
directions. The vertical movement does not require special handling because the local y-axis
is normally chosen to be parallel to the global y-axis. Horizontally, the x- and z-axes usually
coincide with global axes at one support only, and in most other some cases they do not as
can be seen in Fig. 2.10. Therefore, special handling of these boundary conditions through
constraints was necessary.
This is done using a transformation matrix that relates the DOFs in the local x y and the
global coordinate systems X Y . The relationship for the 6-DOF implementation is given
by Equation 2.42.
The transformation for the 7-DOF element is similar except for the additional warping DOF.
Since the physical meaning of warping is two bending moments with same magnitude and
opposite directions as illustrated for example in Fig. 2.11 for an I-section beam, bimoment
has no direction. Therefore, the bimoment in the local coordinate system and that in global
coordinate system are the same.
24
Y glocal
y local
x local
X global
Figure 2.11: Normal stress distribution under warping for a I-section beam
25
u g cos
v
g sin
wg 0
=
xg 0
yg 0
zg 0
sin
cos
cos
sin
0 sin
cos
0
0 u
0 v
0 w
0 x
0 y
1 z
(2.42)
This means that the transformation matrix for the 7-DOF implementation is
u g cos
v
g sin
wg 0
xg = 0
0
yg
zg 0
xg 0
sin
cos
0
0
0
0
0
0
cos
sin
0 sin
cos
0 0 u
0 0 v
0 0 w
0 0 x
0 0 y
1 0 z
0 1 x
(2.43)
It is clear that both transformation matrices contain many zero terms. The full transformation
matrix (14x14) has more zeros than shown in Eq. 2.43 because more DOFs are completely
uncoupled. Direct execution of the transformation resulted in unnecessary computation and
increased runtime. A sparse-matrix implementation was therefore adopted to reduce running
time and storage requirements.
2.6.4 Implementation using Sparse Matrices
The nature of grillage models for curved bridges requires handling many sparse matrices. A
sparse matrix is a matrix with all but few cells holding zero values. The existence of some
sparse matrices does not affect the overall computational time or memory requirements in
small programs that handle all matrices the same way. However, preliminary runs with this
general implementation showed that more than 90% of the computational time is consumed
by transformations that involve large sparse matrices. To overcome this obstacle, a sparse
matrix implementation was needed.
Since most of the matrix multiplication operations are of zero terms, it is possible to skip
most of these operations by just focusing on the nonzero matrix terms. The best approach is
to use algorithms that handle these matrices as reduced sparse matrices. In addition to the
savings in execution time, sparse implementations also save storage requirements. Following
the work of Horowitz et al. (1993), these sparse matrices were compressed into special data
structures. The following example is chosen to better illustrate the method. In the example
an 8x8 sparse matrix [A] is given as:
26
0 15
0 0
0 0
0 0
0 0
0 0
15 0 0 22
0 11 3
0
0 0 0 6
[A] =
0
0 0 0
91 0 0
0
0 0 28 0
(2.44)
[ ]
The first step in handling this sparse matrix is to convert it into a data structure, A , with a
number of elements equal to the number of non-zero cells in [A] ; i.e. eight elements. Each of
the elements in the new data structure holds three pieces of information; namely row number,
column number and value. The resulting data structure is stored in a sequential manner as in
Eq. 2.45
row # col # value
1
1
15
1
4
22
15
1
6
A = 2
2
11
2
3
3
6
3
4
[ ]
5
6
1
3
(2.45)
91
28
[ ]
[ ]
specific feature in the program. Results obtained from the program were compared to those
obtained from commercially available packages. Also comparisons with closed-formed
solutions were performed for one of the problems. The verification study showed that the
program is capable of capturing the general behavior of curved box girder bridges. It is also
capable of capturing the most important aspects of warping behavior.
2.7.1 Comparison with closed-form solution
This verification problem is for a simply-supported curved beam with span length, L = 40 ,
radius, R = 100 . A single concentrated load is applied at midspan, P =10 lb. A total of 160
elements are used to model the problem. Results obtained from the program were compared
to closed-form solution proposed by Konishi and Komatsu (Nakai et al. 1982). In their
solution, a parameter is defined
=L
GJ
EI
(2.46)
For values larger than 9.0, bimoment values can be accurately estimated using the
following expression
Mw =
PR
* R * tan
2
2
2
(2.47)
in which L is the span length, R is the radius of curvature, and P is the concentrated load
acting at mid-span. The central angle is calculated as L R in radians and the parameter is
given as
=R
GJ
EI
(2.48)
Table 2.1 lists the results as obtained from Eq. 2.47 and from the developed program. Three
cross sectional properties were attempted to test the validity of the program over a wide
range of . It can be seen that the difference between both methods is relatively low,
especially for high values of . The larger difference at lower values of should not be
considered a deficiency because 1) the closed-form solution is not an exact solution but
rather a solution based on several assumptions that are often made to simplify the derivation,
2) it will be seen later that values of existing bridges fall in a higher range than the case
considered in this verification study. The conclusion of this study is that the program is
capable of capturing the warping behavior by accurately estimating the bimoment values for
a curved simply supported girder.
28
122
0.1082
0.1118
3.30%
70
0.321
0.325
1.20%
23
2.921
2.682
8.20%
In this verification example, results obtained from the developed program are compared with
ABAQUS results reported in Okeil et al. (2000). The reported results are for the bimoment
distribution of an existing curved box-girder bridge (Bridge 521 of the FDOT inventory see
Chapter 3) under dead loads. They are obtained using the computer program ABAQUS
(1997) which offers a beam element with warping capabilities (Element Type BO31S).
Identical models were built using ABAQUS and the developed program, in which 816
elements where used.
Figures 2.12 and 2.13 show the bimoment distribution along the bridge as obtained from the
program and from ABAQUS, respectively. The figures show that the results are identical
except at the interior supports. At these positions, the bimoment results from the program are
larger that those obtained using ABAQUS. Doubling the element number (using 1632
elements) showed that results were converging, though convergence is not rapid.
2.7.3 Comparison with full shell element models (SAP2000)
also studied. Twenty, forty, and eighty element models were considered. Convergence of
the results was observed for models using less than 80 elements. The results showed that
there is almost no difference between results given by 80-element model and 160-element
model.
Bimoment (N.mm )
3E+09
2E+09
1E+09
0E+00
-1E+09
-2E+09
-3E+09
Figure 2.12: Bimoment distribution along the entire length of Bridge 521 (program
results)
3E+09
Bimoment (N.mm )
2E+09
1E+09
0E+00
-1E+09
-2E+09
-3E+09
Position along span length
Figure 2.13: Bimoment distribution along the entire length of Bridge 521 (ABAQUS
results)
30
Figure 2.14: Dimensions of the cross section for verification example (in inches)
31
The same example was modeled using SAP 2000. Full shell elements with six DOF per node
were used to model the walls of the box as seen in Fig. 2.16. In addition to the fixity of the
supported end of the cantilever, additional intermediate constraints were added to simulate
diaphragms and cross frames to help the cross section maintain its preloading shape. The
lack of these additional constraints would introduce distortion which is not the focus of this
study. Other straining actions (bimoment, bending moment, etc.) are not affected by the
additional constraints if the deformation is small. This measure was necessary since the
element implemented in the program does not account for any distortional deformations. It is
also known that current code provisions require the use of cross frames, which if provided
appropriately, will drop any distortional effects to negligible levels. The positions of
diaphragms are shown in Fig. 2.17. The spacing between adjacent diaphragms is 4.0 inches,
but in the area close to the fixed end, where the normal stress caused by bimoment is more
significant than the normal stress caused by bending moment, more diaphragms are added to
resist distortion, the spacing of diaphragms is reduced to 0.5 inch. Between adjacent
diaphragms which has spacing 4.0 inches, the wall is discretized transversely into 8 elements
and longitudinally into 8 elements. So the element size is 0.25x0.5. Where the spacing of
diaphragms is 0.5 inch, the wall is discretized transversly into 8 elements and longitudinally
into 2 elements, 4 elements, 8 elements respectively. The corresponding element sizes are
0.25x0.25, 0.25x0.125, 0.25x0.0625, respectively. No more refined mesh is discussed
here because of the tolerance in SAP2000 and also, when aspect ratio of shell element is
larger than 4, the accuracy of results will be an issue. The 3 models are designated A, B, and
C respectively. Table 2.2 lists the element size and element number for the three models.
32
Model
Element Size
0.25x0.25
Total Element
Number
2816
0.25x0.5
B
0.25x0.125
3328
0.25x0.5
C
0.25x0.0625
4352
0.25x0.5
33
Figures 2.18 and 2.19 show that the developed program yields deformation results that are
almost identical to those obtained from all 3 shell models.
For a square section subjected to both bending moment and bimoment, the maximum normal
stress happens at the corner. The normal stress is calculated from the bimoment and bending
moment extracted from the computer program as follows:
My
Mx
M
y+
x+
Ix
Iy
I
(2.49)
Results from the SAP2000 model are processed as follows. The top and bottom nodal
stresses are averaged to obtain the membrane nodal stress. The final nodal stress considered
in the verification study is the mean of the average nodal stresses obtained from each element
attached to the node in question. Table 2.3 lists the mean normal stresses at different
positions along the length of models A, B, C, and the results are plotted in Figs. 2.20 and
2.21. Figure 2.21 is an enlarged plot of the same distribution shown in Fig. 2.20 for the last 5
inches of the beam.
The results show that the normal stresses from the developed program and from SAP2000
models are in good agreement in most locations except for the region in the vicinity of the
fixed end of the cantilever beam. In this small region (about 1/20 of the total length), stress
gradient are very high, and Models A, B, and C are not sufficiently refined to pick up the
severe stress changes that occur. The large stress gradient is largely due to the sudden spike
in bimoment near the fixed end, where the bimoment reaches its maximum value.
In summary, the comparisons between SAP2000 and the program indicate that the program is
quite accurate in predicting the behavior of curved box girders. The quality of the results is
naturally dependent on how fine the used mesh is. However, it is obvious that the finite
element is more sensitive to the element size than is the proposed model because of the
spikes in the bimoment distribution that are harder to capture with a shell model. Based on
this verification studies described, it is believed that the program is acceptable and deemed
accurate for the purpose of investigating box girder bridges.
34
0.0016
displacement
0.0014
Program
Model A
Model B
Model C
0.0012
0.001
0.0008
0.0006
0.0004
0.0002
0
0
10
15
20
25
30
35
40
rotation angle(degrees)
0.02
0.018
0.016
0.014
0.012
0.01
0.008
0.006
0.004
0.002
0
Program
Model A
Model B
Model C
0
10
20
30
35
40
50
Table 2.3: Maximum normal stress from program and Models A, B, and C
Distance to fixed end
Program Model A Model B Model C
(inch)
0.000
245.0
100.0
110.0
122.0
0.125
181.0
N/A
94.2
93.0
0.250
116.0
75.0
73.0
71.0
0.500
54.0
52.5
51.9
51.3
0.750
26.0
35.0
35.0
35.5
1.000
14.0
31.0
31.0
31.7
1.250
8.0
23.0
21.6
22.0
1.500
5.4
19.5
20.0
20.0
1.750
4.0
14.5
14.0
13.9
2.000
3.0
13.7
14.6
15.0
4.000
6.0
8.0
8.0
8.5
6.000
9.0
10.0
10.3
10.3
8.000
12.0
13.4
13.4
13.1
10.000
15.0
15.5
15.5
15.6
12.000
18.0
20.5
20.2
20.1
14.000
21.0
21.7
22.0
22.0
16.000
24.0
26.7
26.7
26.8
18.000
27.0
28.0
56.4
24.0
20.000
30.0
33.5
33.4
33.4
22.000
33.0
34.3
34.2
34.2
24.000
36.0
40.0
40.0
40.1
26.000
39.0
40.5
40.3
40.3
28.000
42.0
47.3
46.6
46.6
30.000
45.0
46.5
46.4
46.4
32.000
47.0
53.0
53.0
52.7
34.000
52.0
52.3
52.3
52.2
36.000
53.0
59.3
59.2
59.3
38.000
56.0
55.2
55.2
55.2
36
300
Program
Model A
Model B
Model C
250
200
150
100
50
0
0
10
20
30
40
Figure 2.20: Maximum normal stress comparison along entire beam length
300
Program
Model A
Model B
Model C
250
200
150
100
50
0
0
Figure 2.21: Maximum normal stress comparison in the vicinity of fixed end
37
3.1 Background
This chapter presents a detailed investigation of warping related stresses in eighteen existing
box girder bridges chosen from the Florida Department of Transportation inventory. The
bridges are carefully selected to cover a wide range of design parameters including horizontal
curvature, cross sectional properties, and number of spans. They were designed by different
firms and were constructed at different times and are considered to be representative of
current design practice. Forces are evaluated from analyses that account for the construction
sequence and the effect of warping. Loading is considered following the 1998 AASHTOLRFD provisions. The differences between stresses obtained taking into account warping
and those calculated by ignoring warping are used to evaluate the effect of warping.
Analysis results show that warping has little effect on both shear and normal stresses in all
bridges. Current design provisions are discussed in light of the analysis results.
particular section only (i.e. during construction). This guidance is adapted from the work of
Nakai and Heins (1977), who investigated a variety of curved bridge types and proposed
criteria based on cross-section properties, bridge length, and subtended angle that would
allow engineers to determine when warping is significant. The study by Nakai and Heins
(1977) has several limitations including: 1) idealized loading and boundary conditions were
assumed; 2) although, normal stresses due to warping were considered, shear stresses due to
warping were ignored; and 3) the effects of centrifugal forces were not accounted for.
Centrifugal forces always occur in curved bridges and can be substantial.
Other researchers have also tried to quantify the significance of warping in curved box
girders. Trukstra and Fam (1978) investigated the effect of diaphragms on the behavior of
curved box girder bridges. They conducted a parametric study using finite element models
and investigated for simple load cases the ratio between stresses calculated from the finite
element model and corresponding stresses obtained from idealized beam models. Both
concrete and composite single box girders were considered. As expected, the results showed
that diaphragms improve load distribution and positively influence stress ratios. A wide
range of radii was covered in the study. Girders that showed large stress ratios belong to a
group with small radii, which fall into an impractical range (R<30.48m [100 ft]). Shear
stresses were also not investigated in this study.
More recently, Waldron (1988) investigated the effect of warping on normal stresses in
single box girders. Forces were calculated by deriving closed form solutions of the
fundamental equation governing torsion and warping for special loading cases. Using
concrete box examples, it was shown that warping could increase normal stresses by as much
as 29%. This high stress ratio corresponds to a theoretical loading condition where a single
concentrated load acts on one of the webs at midspan. For truck loads (following the British
code), stress ratios drop to around 5%. Based on the study, it was concluded that the widthto-depth ratio significantly impacts the normal warping stress ratio.
The studies summarized above all suffer from a number of common drawbacks. They did not
address warping shear stresses and were based on idealized loading and boundary conditions.
They also did not address the construction sequence of composite box girders. This study
addresses all of these issues and presents a detailed investigation of warping stresses in
curved composite box girders. The study is conducted on 18 bridges from the Florida
Department of Transportation (FDOT) inventory. The bridges are carefully selected to cover
a wide range of design parameters including horizontal curvature, cross sectional properties,
and number of spans. The bridges were designed by different firms and were constructed at
different times and are considered to be representative of current design practice. A summary
of the main properties of the bridges is given in Table 3.1 and further information can be
found in Okeil et al. (2000) and Okeil and El-Tawil (2002). The 1998 AASHTO-LRFD load
provisions are considered in the study and the effect of warping on both normal and shear
stresses is quantified. The implications of the analysis results with respect to current design
provisions are then discussed.
39
Spans
Radius of
curvature
Min
(m)
Max
(m)
Min
(m)
Max
(m)
Finite
element
length
(mm)
Lanes
Number
of finite
elements
Number
of load
cases
390
54.86
71.17
188.98
188.98
152.4
1923
74
521
23.16
36.58
1758.90
1758.90
101.6
816
35
525
41.54
49.59
Straight
Straight
101.6
897
37
528
48.16
58.52
1746.38
1746.38
152.4
1658
68
537
30.48
64.92
218.30
218.30
152.4
1390
58
538a
37.19
52.43
436.60
3033.17
152.4
1585
65
538b
46.33
64.31
436.60
873.19
152.4
1762
69
538c
50.90
64.01
873.19
873.19
152.4
1508
61
538d
36.88
64.01
436.60
873.19
152.4
2486
100
539
39.62
55.93
218.60
218.60
152.4
1539
63
540
23.77
54.25
290.76
290.76
152.4
1661
69
541a
38.25
63.09
250.40
431.51
152.4
1686
69
541b
32.16
61.26
250.40
645.52
152.4
1604
60
542a
43.89
62.18
349.58
349.03
152.4
2096
84
542b
34.75
59.13
349.58
8732.18
152.4
1834
74
598
44.35
52.27
107.56
211.68
152.4
1895
51
606
56.08
80.37
438.88
875.00
101.6
1310
69
607
46.94
66.85
870.91
Straight
101.6
1582
60
40
E
ab
3
G d
2 a3
+
Fd 3 Fo
(3.1)
where E is the modulus of elasticity, G is the shear modulus, a the spacing between cross
frames, b is the distance between flanges, d is the length of bracing member, Fo is the area
of top flange, and Fd is the area of bracing member.
After hardening, the concrete deck becomes an integral part of the cross section and its
contribution is accounted for. The equivalent plates contribution becomes small compared
to the deck, and is ignored. In accounting for the concrete deck, it is first transformed into an
equivalent steel plate using an appropriate modular ratio; E c E s . This transformation is
justified when the deck is completely in compression, but is approximate when the deck or
parts of it are under tension. Since 1) it is not clear when the deck will be completely in
tension because of the combined effect of moment and torsion, and 2) the contribution of the
top bracing is being ignored, the use of the uncracked properties is deemed reasonable in this
work. Other researchers have also made use of this approximation in the past including
Johnson and Mattock (1967), whose work forms the basis of several provisions in AASHTO
LRFD (1998). Figures 3.1 and 3.2 show a few of the geometric functions for one of the
bridges considered before and after integration of the deck in the cross section.
41
2133.6
top bracing
1143
558.8
1143
558.8
(b) S x
(c) S y
(d)
(e) S
42
2133.6
203.2
6502.4
1549.4
1143
558.8
1143
1549.4
558.8
(b) S x
(c) S y
(d)
(e) S
Figure 3.2: Geometric properties of closed cross section (Dimensions in mm, in mm2,
S x S y in mm3, S in mm4)
43
For simple geometries (regular spans, curvatures, etc.) and loadings (uniform torque,
single concentrated torque), girder forces are a function of a dimensionless parameter, ,
which is given as (Nakai and Yoo 1988)
=L
GK
EI
(3.2)
where L is the span length, G is the shear modulus, K is the torsional constant, E is the
modulus of elasticity, and I is the warping constant. A large implies that the
contribution of warping to stiffness is small and that warping related stresses are therefore
low. On the other hand, a small implies that the warping contribution to stiffness is large
and that warping related stresses could be high.
The reliance of member forces on is true only for certain idealized loading, geometric, and
boundary conditions. In real bridges, the cross-section and radius of horizontal curvature both
vary along each span as well as from span to span, which renders the use of Eq. 3.2
impossible. To account for these variations, a weighted (calculated for each span) is used
as follows. The relationship between this weighted parameter and warping stresses is
discussed later on.
=
L
GK
dL
EI
(3.3)
Two values are computed for each span of each bridge corresponding to quasi-closed and
closed conditions. Figures 3.3 show a plot of these values versus the average L / R ratio
for corresponding spans. The values are calculated for spans where the critical normal
stresses (Fig. 3.3a) and shear stresses (Fig. 3.3b) take place as described later. As expected,
Fig. 3.3 clearly shows that closed cross-sections possess better torsional qualities than open
cross-sections, i.e. higher values. Although there is some scatter, it is clear that current
design practices yield values with an average of about 38 for quasi-closed cross sections
and 114 for closed cross sections. A slight increase in the trend of is also observed with
higher L / R , which reflects the more efficient designs for spans with a sharper radius of
curvature.
44
180
160
140
120
100
80
60
40
20
0
0
0.1
0.2
0.3
quasi-closed
0.4
closed
L/R
0.5
(a)
160
140
120
100
80
60
40
20
0
0
0.1
0.2
quasi-closed
0.3
closed
0.4 L/R
0.5
(b)
Figure 3.3: Geometric properties of closed cross section (Dimensions in mm, in mm2,
S x S y in mm3, S in mm4)
45
Detailed descriptions of the 18 bridges considered in this study can be found in Okeil et al.
(2000). For completeness, a summary of the modeling details is provided next.
The analyses are conducted using the computer program ABAQUS (1997). All 18 girders
considered in this study are modeled using a three-dimensional beam-column element
(BO31S) that accounts for warping . The element has 7 degrees of freedom per node, and is
similar to the element described in Chapter 2. Although several of the bridges have more
than one box, only one box girder is modeled per bridge to reduce the modeling and
computational effort. Each box girder is assumed to support a slab that is wide enough to
accommodate at least one traffic lane. This implies a load distribution factor of at least 1.0
lane per girder, which is conservative when compared to distribution factors from AASHTOLRFD specifications (1998) for straight girders.
Small elements (101.6mm [4in] or 152.4mm [6in]) are used to model the bridges resulting in
hundreds of elements per girder. This is necessary in order to accurately calculate the
warping related forces. For example, the bimoment experiences large spikes at supports,
which are difficult to capture unless small elements are used. Verification studies confirmed
that the chosen element sizes are sufficient to provide reasonable accuracy. Intermediate
support locations are restrained against vertical and transverse translations, and twisting
rotations. The warping degree of freedom is not restrained anywhere because diaphragms are
not capable of providing significant resistance to out of plane deformations. However,
continuity over intermediate supports leads to the development of bimoments and torsional
warping moments, which is not the case at end supports where the developed bimoment is
zero. Table 3.1 lists the number elements used for each bridge, which ranged from 816 to
2486. The table also lists other details for the bridges covered in this study.
3.3.3 Bridge loading and resulting forces
Dead loads are estimated based on the dimensions and material properties given in blueprints
provided by FDOT. Live loads are considered according to AASHTO-LRFD (1998).
Following the HL-93 loading, a uniform lane load of 9.3 N/mm, uniformly distributed load
over a 3000 mm width, is considered in addition to a tandem load (two 110kN axles).
Transversely, the loads are positioned at the outermost possible location to generate the
maximum torsional effects (see Fig. 3.4). The centrifugal force (CF) is taken into account for
curved segments of the bridges. From these loads, the dead load force distributions and
envelopes for the live load force distributions are generated. Each set included, M x , M y ,
M , V x , V y , T , Ts , and T , which are described in the next section. Figure 3.5 shows an
example of the live load factored (Strength I limit state) envelopes for warping related forces
obtained from the analysis of an idealized 3-span pilot bridge.
46
9.3 N/mm
3000
Lane Load
110 kN
CE
1800
1800
600
b
Figure 3.4: Positioning of live loads for a single lane bridge (dimension in mm)
Since ABAQUS does not provide results for the warping torsional moment, T , the fact that
it is the derivative of the bimoment (i.e. T = M ) is used to obtain these results. A
numerical differentiation scheme (Greenspan and Casulli 1988) of the bimoment, M , is
utilized. To enhance the quality of the numerical results, three points are used to numerically
differentiate M at any point; the point under consideration and the points before and after.
For example, determining the warping torsional moment at node i involves the values of the
warping moment at the nodes i 1 and i + 1 as well as the elements length, Lelem
(T )i
(M )i 1 + (M )i +1
2Lelem
(3.4)
47
At points where abrupt changes take place such as supports, a forward (point under
consideration and two following points), or a backward (point under consideration and two
previous points), differentiation scheme are used so that the spikes can be accurately
captured. Equations 3.5 and 3.6 Give the expressions for the forward and backward
numerical differentiation, respectively.
(T )i
(T )i
1st Span
2E+8
3(M )i + 4(M )i +1 + (M )i + 2
2 Lelem
(3.5)
(M )i 2 4(M )i 1 + 3(M )i
(3.6)
2 Lelem
2nd Span
1st Span
3rd Span
Bimoment
3E+5
2nd Span
3rd Span
2E+5
1E+8
kN-mm
kN-mm2
1E+5
0E+0
0E+0
-1E+5
-2E+5
-3E+5
-1E+8
6E+5
4E+5
4E+5
2E+5
2E+5
kN-mm
kN-mm
6E+5
0E+0
0E+0
-2E+5
-2E+5
-4E+5
-4E+5
-6E+5
-6E+5
Figure 3.5: Envelope of warping-related straining actions due to live loads for idealized
bridge
48
Normal stresses including the effect of warping - are calculated as follows (Nakai and Yoo
1988):
M
Mx
M
M
y + y x + = approx +
Ix
Iy
I
I
exact =
(3.7)
where I x and I y are the moments of inertia about the x- and y-axes, and x and y are the
distances from the centroid of the cross section. Equation 3.7 shows that the exact normal
stress, exact , is generated by the bending moments ( M x , M y ) and bimoment ( M ) which
causes warping. The sum of the first two terms is from classical beam theory, which does not
account for warping, and will be referred to as approx . The third term is a function of the
warping constant, I , and the warping function, .
The approximate shear stress, approx , is calculated from classical beam theory (i.e. by
ignoring the effects of warping) using the following equation:
approx =
Vy
tI x
S x (s) +
Vx
T
S y ( s) +
2tAc
tI y
(3.8)
where V x and V y are the shear forces and T is the torsional moment acting on the cross
section. The other terms in the expression represent the geometric properties of the closed
cross section; namely, wall thickness ( t ), moments of area ( S x (s ) and S y (s ) ), and enclosed
cross-sectional area ( Ac ). The third term assumes that the entire torsional moment is St.
Venant torsion. If warping is considered, the torsional moment must be split into its two
constituent terms (pure torsion, Ts , and warping torsion, T ), and the shear stresses can be
calculated as follows:
exact =
Vy
tI x
S x ( s) +
Vx
T
T
S y ( s ) + s + S ( s )
tI y
2tAc tI
(3.9)
The geometric properties I (warping constant) and S (s ) (sectoral area) are used to
calculate the shear stress due to warping torsional moments (last term in Eq. 3.6).
49
exact approx
approx
(3.10)
WSR S =
exact approx
approx
(3.11)
50
DL , exact =
LL , exact =
Stress Ratios
M yDL
M xDL
y+
x
I yO
I xO
M DL
I O
M yDL
M xDL
M
y+
x + DL
I xO
I yO
I O
DL , approx =
M yLL
M xLL
M
y+
x + LL
I xC
I yC
I C
LL , approx =
M yLL
M xLL
y+
x
I xC
I yC
=
WSR DL
=
WSRLL
LL, exact =
V yDL
tI xO
V yLL
tI xC
V xDL
T
T
S yO ( s) + sDL + DL S O
tI yO
2tAcO tI O
DL , approx =
T
T
V xLL
S yC ( s) + sLL + LL S C ( s
2tAcC tI C
tI yC
LL , approx =
S xO ( s ) +
S xC ( s) +
V yDL
tI xO
V yLL
tI xC
LL , approx
T , approx
S xO ( s ) +
V xDL
T
S yO ( s ) + D
tI yO
2tA
WSRDL
=
S xC ( s ) +
V xLL
T
S yC ( s ) + LL
tI yC
2tAc
WSRLL
=
LL , approx LL , exact
LL , approx
51
M LL
I C
M DL
M
+ LL
I O
I C
WSRT =
DL, exact =
DL , approx
WSRT =
T , approx T , exact
T , approx
3.5 Results
The WSR for all cases summarized in Table 3.2 are calculated using the procedures
described earlier. Two sets of plots are generated in an attempt to identify trends in the
results. Figures 3.7 and 3.8 show the WSR for normal and shear stresses plotted versus the
average L/R ratios for corresponding spans. The WSR are also plotted versus , as shown in
Figs. 3.9 and 3.10. Each of the four figures provides three plots corresponding to the ratios
due to dead, live and total (combined dead and live) loads. The dashed line in each plot
represents the average level. In spite of significant scatter in some of the plots, several
observations are evident from Figures 3.7 through 3.10. These observations are discussed
next.
3.5.1 Normal stresses
During construction (i.e. quasi-closed cross-section under dead load), ignoring warping
implies that normal stresses are underestimated by an average of 0.84%. The most critical
case is underestimated by 1.90%. In designing for live loads (closed section), the average
effect of warping is an additional 0.25% with the most severe case being 1.88%. The
average WSR for total stress effect (closed section subjected to dead and live loads) is 0.34%
with a maximum of 3.16%. In evaluating these numbers, readers should keep in mind that
the locations at which the maximum WSR for dead, live, and combined loading are different.
Although the plots in Figs. 10 and 12 do not show any conclusive trends because of scatter in
the data, it is clear that the magnitude of the WSR are quite small (less than 3.16% for all
cases).
3.5.2 Shear stresses
Although the WSR s pertaining to shear are somewhat higher than those corresponding to
normal stresses, the ratios are still relatively low. The calculations show that the average dead
load shear stresses are underestimated by 1.57% with a maximum of 7.42%. After casting
the concrete deck, the cross sections properties are greatly enhanced, and ignoring warping
is actually on the conservative side with a tendency to overestimate stresses by an average of
1.52%. For combined loading (Fig. 3.10 (c)), total stresses are underestimated if warping is
not taken into account by an average of 1.37%.
3.5.3 Effect of
figures is scattered and does not appear to have a specific trend. There are several reasons
that can be put forward to explain this: 1) although spans with a larger are torsionally
stronger, they are probably subjected to larger demands; 2) the bridges were designed by
different firms and hence the level of conservatism in design as well as the design models
and tools vary from one bridge to another; 3) calculated according to Eq. 3.3 is a weighted
value that may not accurately reflect the vulnerability of a critical cross-section to warping as
well as calculated from Eq. 3.2 would for idealized conditions.
53
2.0
1.5
WSRDL (%)
AASHTO limit
1.0
0.5
0.0
0
0.1
0.2
L/R
0.3
0.4
1.5
WSR LL
(%)
1.0
0.5
0.0
0
0.1
0.2
L/R
0.3
0.4
0.1
0.2
L/R
0.3
54
0.4
9.0
AASHTO limits
6.0
WSR DL
(%)
3.0
0.0
-3.0
0
0.1
0.2
0.3
0.4
0.5
0.3
0.4
0.5
L/R
0.1
0.2
L/R
0.0
-2.0
(%) -4.0
WSR LL
AASHTO limits
-6.0
-8.0
0.1
0.2
L/R
0.3
0.4
0.0
-1.0
WSR T (%)
-2.0
-3.0
-4.0
AASHTO limits
-5.0
-6.0
55
0.5
2.0
1.5
WSRDL
(%)
1.0
0.5
0.0
0
10
20
30
40
50
60
WSR LL
(%) 1.0
0.5
0.0
0
50
100
150
200
150
200
50
100
56
9.0
6.0
WSR DL
(%) 3.0
0.0
-3.0
0
10
20
30
40
50
60
70
50
100
150
200
0.0
-2.0
WSR LL (%)
-4.0
-6.0
-8.0
50
100
0.0
-2.0
-3.0
-4.0
-5.0
-6.0
57
150
58
4.1 Introduction
This chapter describes an investigation of the live load distribution factor (DF). The live load
distribution factor is an integral part of current AASHTO design codes. AASHTO LRFD
(1998) provides distribution factors for several structural systems including composite box
girder bridges. The suggested DFs are limited to straight bridges and bridges with slight
curvatures and are based on the work of Ho (1972). The AASHTO Curved (1997) does not
provide specifications regarding DFs for composite curved box girder bridges. However,
the commentary of AASHTO Curved (1997) provides alternate expressions for DFs that can
be used for composite curved box girder bridges. The provided expressions are based on the
work of Ho and Reilly (1971) and Heins (1972). The expressions in the commentary
consider three variables only in determining the DFs; namely number of girders, G , girder
spacing, S , and the number of lanes, N . Equations 4.1 and 4.2 show these expressions for
bending and torsional moments.
gM =
gT =
S
2 .6 + 0 .6
(4.1)
GS
N
(4.2)
GS
15.3 + 1.5
N
Since the previous expressions are derived based on analyses of straight girder, curvature still
needs to be accounted for. This is done through modifiers following Eq. 4.3 and 4.4 for
bending and torsional moments, respectively:
M = 1.0 +
T = C +
A B
+
R R2
(4.3)
D
R2
(4.4)
59
This simplification of a very complex behavior leads to ignoring other parameters such as the
ratio between the lateral and longitudinal stiffness of the system. Furthermore, the torsional
stiffness of each box is not accounted for. It should also be noted that the studies based on
which the current DFs were derived, were conducted decades ago with the available tools at
the time. Current advancements allow more in-depth studies with more sophisticated models
than were available before. Such shortcomings led to a call by the National Cooperative
Highway Research Program (NCHRP) to revisit the current expressions for live-load DFs
for all types of bridges. This part of the study serves as a good preliminary investigation to
the research program announced by NCHRP.
V girder = gV V simplified
(4.5)
The expressions for DFs are usually obtained through analyses similar to those described
here, except that they are done in reverse order; i.e. the distribution factor is obtained by
dividing results obtained from analyzing the real system by those of the simplified system. A
regression of results leads to the expression of the DFs. The main benefit of using the DF is
that it is possible to determine internal forces in a three-dimensional (3-D) system such as a
bridge by analyzing a simplified single girder system.
element. In choosing for the study at hand, the advantages and disadvantages of both
approaches were weighed in. The findings are summarized next
4.3.1 Shell Models vs. Grillage Models
Modeling the entire system using shell elements is undoubtedly more involved and provides
additional information that other models are not capable of providing. However, shell
elements provide results on a stress or stress resultant level, which introduces the following
shortcomings. First, current codes focus more on the member level rather than on the
material level. This is because the LRFD approach requires determining the strength of the
member being designed in contrast with the allowable stress design method (ASD), which
sets an allowable stress at the material level. Converting stresses obtained from shell model
analyses to section forces is possible by integration of the appropriate stresses over the
desired cross section. The integration process is tedious and time consuming. Second, the
quality of stresses or stress resultants obtained from shell models are sensitive to various
factors such as stress concentrations due to shear lag or any other sort of disruptions in the
cross section such as stiffeners. Stresses are also affected by the type of element being used
in the analysis, which in the opinion of the authors has an effect on how reliable the resulting
DFs are.
Grillage models are more straightforward and provide direct results in the form of internal
forces and moments. The simplicity of grillage models comes at the cost of losing some of
the details in modeling three-dimensional objects, especially box girders systems. This could
lead to inaccurate estimates of the DFs. However, many of the details that cause such
discrepancies can be adequately modeled using grillage models. For example, an analysis of
a box girder bridge using linear elements to model girders as well as the concrete deck leads
to inaccurate results because the span of the deck is exaggerated as can be seen in Fig. 4.1.
In the figure, the deck spans the distance between the box upper flanges. However, linear
elements passing through the centroid implies a deck system as shown in the lower system.
centroid
Figure 4.1: Grillage modeling and its effect on deck span length
61
As will be seen later, it is possible to overcome such drawbacks of grillage models through
modeling adjustments. Accordingly, the choice of a shell model was deemed not necessary
and a sophisticated grillage model was used for the DF study.
4.3.2 Determination of Distribution Factors
For each bridge configuration, two bridge models were built. The first model is a single
girder model, which represents what the designer would normally use in every day design.
For engineers, the analysis of such a model is quite simple and straightforward with readily
available analysis tools. Figure 4.2 shows a schematic of one of the single girder models
loaded with a group of concentrated loads representing a standard truck (HS-20) loading.
Figure 4.2: Schematic of single girder models used in determining the distribution
factor
The second model is a detailed grillage model of the bridge system. Linear elements were
used to model the bridge components such as girders, deck, and diaphragms. Figure 4.3
shows a schematic of the node and element arrangements for one of these models having
three box girders. As it appears from the figure, girder elements (bold lines) connect nodes
that are not shared with deck elements. Deck elements connect nodes according to deck
dimensions, however avoiding direct contact with girder nodes. The connectivity of the deck
and the girders is achieved through rigid links that apply constraints between nodes at their
ends. The constraints are imposed in a master/slave type of relationship in such a way that
achieves a continuity condition similar to that of the real structure. This arrangement avoids
overestimating deck span lengths as described earlier.
The grillage model is verified by comparing results to those obtained using folded plate
theory for a straight three-cell box-girder bridge (Johnson et al. 1967). The verification
example is provided with only two diaphragms at the supports, thus the combined torsional
rigidity method is used to model distortional behavior. The combined torsional rigidity
62
deck
elements
deck
node
girder
node
girder
element
rigid
links
Figure 4.3: Schematic of grillage models used in determining the distribution factor
d/2
d/2
Figure 4.4: Dimensions for calculation of total torsional constant based on BEF
63
The flexural rigidity of the top flange Da, bottom flange Db, web Dc is first
determined:
D=
Et 3
12(1 2 )
(4.6)
The out-of-plane shear in the bottom flange per unit torsional load is then
calculated:
3
1
[(2a + b )abc] + ba
Dc
Da
3
3
(a + b ) a + 2c (a 2 + ab + b 2 ) + b
Db
Da Dc
(4.7)
The vertical deflection of one web per unit torsional load can now be calculated
as:
2
c 2ab
ab
a b
(
)
2
a
b
+
24(a + b ) Dc a + b
Da a + b
1
=
EI c
(4.8)
0.25
(4.9)
C=
a 2 I c ( l )
l 2 1.6w
(4.10)
64
1
1 1
= +
Jt C J
(4.11)
in which J is the classic torsional constant for closed cross sections (see
Appendix A).
deflection/average deflection
The three-girder bridge example reported in Johnson et al. (1967) was modeled using the
suggested grillage model after obtaining the torsional constant following the above
procedure. Figure 4.5 Show a comparison between results obtained from the suggested
model and those reported by Johnson et al. The results are for the deflections of the deck at
mid-span of the bridge due to an eccentrically positioned standard truck acting at midspan.
As expected, the lateral flexibility leads to unequal deflections. Figure 4.5 shows that
deflections under the loads are about 2.5 times the average deflection. It can be seen that the
grillage model predicts deflections that are in good agreement with those reported by
Johnson et al. (1967). Experimental results obtained from a 1:50 scale model were in good
agreement with the same results.
0.5
0
-0.5
-1
-1.5
-2
-2.5
0
100
200
300
400
The single girder and grillage models are used in the subsequent analyses which are
performed in three stages:
-
Stage I: A single girder model is built for each of the bridges in the parametric
study. The model is then analyzed by loading it with AASHTO-LRFD (1998)
loads for a full lane width (HL-93) by sweeping the truck along the longitudinal
axis of the model. These analyses will yield the maximum force effects (bending
moments, shear forces, and torsional moments) in the single girder bridge model,
65
single . The maximum forces are recorded for later use in determining the
distribution factor. Truck locations causing these maximum forces are also
stored. These locations serve as a starting point for Stage II analyses.
-
Stage II: A grillage model is built for each of the bridges. Analyses of the
grillage model are conducted for the same loading (HL-93) as linear models,
however, the number of lanes loaded is based on the actual roadway width. The
truck positioning in the longitudinal direction is predetermined from the single
girder analyses in Stage I. The transverse location causing the maximum force
effects is determined by sweeping the whole loading of the lane (truck + lane)
laterally. The maximum forces caused by the most severe positioning of loads
are recorded for all girders (exterior and interior), girder .
Stage III: The distribution factors are determined according to Eq. 4.12 based on
the results obtained in Stages I and II.
g =
girder
single
(4.12)
66
Number of
Lanes, N L
Number of
Girders, N
Girder
Spacing, S
Span Length,
Radius of
Curvature, R
L
(m)
(m)
(m)
Number of
Spans
4.10
2.74
5.90
25
100
3.94
50
250
2.95
100
500
5.14
3.85
3.08
67
R
100
100
100
100
100
100
100
100
100
100
100
100
100
250
250
250
250
250
250
250
250
250
250
250
250
250
500
500
500
500
500
500
500
500
500
500
500
500
Spans
2
2
2
2
2
2
3
3
3
4
4
4
4
2
2
2
2
2
3
3
3
3
4
4
4
4
2
2
2
2
2
3
3
3
4
4
4
4
NL
2
3
3
4
4
4
4
4
4
3
4
4
4
2
3
4
4
4
3
4
4
4
3
4
4
4
2
3
3
4
4
3
4
4
3
4
4
4
3
3
4
3
4
5
3
4
5
3
3
4
5
3
4
3
4
5
3
3
4
5
3
3
4
5
3
3
4
4
5
3
4
5
3
3
4
5
g Me +
gM
Eq. 4.1 Prog
3.60
3.60
3.60
5.14
3.60
3.08
5.14
3.60
3.08
3.60
5.14
3.60
3.08
3.60
3.60
5.14
3.60
3.08
3.6
5.14
3.6
3.08
3.6
5.14
3.6
3.08
3.6
3.6
3.6
3.6
3.08
3.6
3.6
3.08
3.6
5.14
3.6
3.08
0.735
1.316
0.862
1.690
1.316
1.015
1.690
1.316
1.015
1.316
1.690
1.316
1.015
0.735
0.862
1.690
1.316
1.015
1.316
1.690
1.316
1.015
1.316
1.690
1.316
1.015
0.735
1.316
0.862
1.316
1.015
1.316
1.316
1.015
1.316
1.690
1.316
1.015
0.942
1.149
1.093
1.562
1.200
0.937
1.563
1.202
0.938
1.151
1.563
1.202
0.938
0.933
1.074
1.532
1.173
0.907
1.133
1.534
1.175
0.908
1.133
1.534
1.175
0.908
0.929
1.125
1.067
1.163
0.897
1.126
1.165
0.898
1.027
1.524
1.165
0.898
g Mi +
g Me
g Mi
Diff. (%)
Prog
Diff. (%)
Prog
Diff. (%)
Prog
Diff. (%)
-21.9
14.6
-21.1
8.2
9.7
8.3
8.1
9.5
8.2
14.4
8.1
9.5
8.2
-21.2
-19.7
10.3
12.2
11.9
16.2
10.2
12.1
11.8
16.2
10.2
12.1
11.8
-20.8
17.0
-19.2
13.1
13.2
16.9
13.0
13.1
28.2
10.9
13.0
13.1
0.777
1.064
0.952
1.509
1.119
0.942
1.511
1.121
0.943
1.064
1.511
1.121
0.943
0.776
0.950
1.508
1.110
0.928
1.065
1.510
1.112
0.929
1.065
1.510
1.112
0.930
0.776
1.064
0.948
1.107
0.923
1.065
1.108
0.925
1.065
1.510
1.109
0.925
-5.3
23.8
-9.4
12.0
17.6
7.8
11.9
17.5
7.6
23.7
11.9
17.5
7.6
-5.3
-9.2
12.1
18.5
9.3
23.6
11.9
18.4
9.2
23.6
11.9
18.4
9.2
-5.2
23.7
-9.1
18.9
9.9
23.6
18.8
9.8
23.6
11.9
18.7
9.7
0.970
1.167
1.125
1.575
1.217
0.942
1.571
1.216
0.939
1.167
1.571
1.216
0.939
0.969
1.116
1.551
1.194
0.920
1.153
1.548
1.193
0.917
1.154
1.549
1.193
0.917
0.968
1.147
1.113
1.185
0.912
1.148
1.185
0.910
1.149
1.541
1.185
0.910
-24.2
12.8
-23.4
7.3
8.1
7.7
7.6
8.3
8.1
12.8
7.6
8.3
8.1
-24.1
-22.8
9.0
10.3
10.3
14.1
9.2
10.4
10.7
14.1
9.1
10.3
10.6
-24.0
14.7
-22.5
11.0
11.3
14.6
11.1
11.6
14.5
9.7
11.0
11.5
0.759
1.051
0.963
1.516
1.128
0.947
1.523
1.129
0.948
1.055
1.523
1.129
0.948
0.759
0.964
1.516
1.124
0.941
1.055
1.522
1.125
0.941
1.055
1.523
1.125
0.942
0.759
1.052
0.964
1.122
0.938
1.055
1.123
0.939
1.056
1.523
1.124
0.940
-3.1
25.2
-10.5
11.5
16.7
7.1
11.0
16.6
7.1
24.8
11.0
16.6
7.0
-3.2
-10.5
11.5
17.1
7.9
24.8
11.1
17.0
7.8
24.7
11.0
17.0
7.7
-3.1
25.2
-10.5
17.3
8.2
24.7
17.2
8.1
24.7
11.0
17.1
8.0
68
Spans
NL
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
2
2
2
2
2
2
3
3
3
3
3
3
4
4
4
4
4
4
2
2
2
2
2
2
3
3
3
3
3
3
4
4
4
4
4
4
2
2
2
2
2
2
3
3
3
3
3
3
4
4
4
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
2.74
3.94
2.95
5.14
3.85
3.08
g Me +
gM
Eq. 4.1 Prog
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.819
1.254
0.941
1.690
1.269
1.015
0.734
1.120
0.860
1.508
1.165
0.949
0.735
1.120
0.861
1.509
1.166
0.949
0.735
1.120
0.861
1.509
1.166
0.949
0.711
1.091
0.824
1.472
1.123
0.901
0.710
1.091
0.823
1.472
1.122
0.900
0.710
1.091
0.823
1.472
1.122
0.900
0.700
1.078
0.809
1.457
1.105
0.881
0.700
1.079
0.809
1.458
1.106
0.881
0.700
1.079
0.809
1.458
1.106
0.881
g Mi +
g Me
g Mi
Diff. (%)
Prog
Diff. (%)
Prog
Diff. (%)
Prog
Diff. (%)
11.47
11.95
9.38
12.12
8.90
6.94
11.45
11.90
9.34
12.03
8.84
6.90
11.45
11.89
9.34
12.03
8.84
6.90
15.20
14.93
14.17
14.81
13.00
12.61
15.35
14.95
14.33
14.80
13.04
12.76
15.35
14.94
14.33
14.79
13.03
12.76
16.95
16.26
16.36
16.00
14.82
15.15
16.95
16.19
16.37
15.92
14.75
15.16
16.95
16.18
16.37
15.91
14.74
15.16
0.695
1.060
0.818
1.424
1.101
0.920
0.695
1.060
0.819
1.425
1.102
0.916
0.695
1.060
0.819
1.425
1.102
0.916
0.694
1.059
0.808
1.423
1.088
0.894
0.694
1.060
0.808
1.423
1.088
0.893
0.694
1.060
0.808
1.423
1.088
0.893
0.693
1.058
0.802
1.421
1.081
0.884
0.694
1.060
0.803
1.423
1.082
0.884
0.694
1.060
0.803
1.423
1.082
0.884
17.84
18.31
14.98
18.69
15.17
10.37
17.82
18.26
14.94
18.63
15.12
10.76
17.82
18.26
14.94
18.63
15.11
10.76
17.98
18.33
16.43
18.80
16.65
13.46
17.94
18.28
16.53
18.74
16.65
13.62
17.94
18.27
16.53
18.74
16.65
13.60
18.16
18.46
17.27
18.94
17.38
14.85
17.96
18.28
17.17
18.80
17.24
14.77
17.96
18.28
17.17
18.79
17.24
14.76
0.738
1.121
0.863
1.521
1.164
0.945
0.739
1.123
0.863
1.524
1.165
0.945
0.740
1.121
0.864
1.527
1.167
0.945
0.725
1.104
0.841
1.499
1.137
0.915
0.728
1.111
0.844
1.506
1.142
0.916
0.728
1.111
0.844
1.506
1.142
0.916
0.718
1.097
0.832
1.489
1.126
0.903
0.722
1.104
0.835
1.497
1.131
0.904
0.722
1.104
0.835
1.498
1.132
0.905
10.99
11.85
9.10
11.15
9.02
7.39
10.85
11.60
9.02
10.90
8.89
7.42
10.69
11.87
8.92
10.70
8.72
7.41
12.98
13.54
11.86
12.79
11.56
10.90
12.51
12.86
11.52
12.26
11.12
10.80
12.50
12.83
11.50
12.23
11.09
10.78
13.97
14.28
13.11
13.48
12.62
12.39
13.43
13.55
12.69
12.91
12.12
12.23
13.42
13.52
12.66
12.87
12.08
12.20
0.685
1.046
0.823
1.422
1.114
0.920
0.686
1.049
0.825
1.429
1.117
0.921
0.688
1.054
0.827
1.436
1.120
0.924
0.686
1.047
0.818
1.423
1.109
0.909
0.690
1.057
0.825
1.439
1.118
0.915
0.690
1.057
0.825
1.439
1.118
0.916
0.686
1.047
0.816
1.423
1.107
0.905
0.690
1.057
0.823
1.439
1.116
0.912
0.690
1.058
0.823
1.439
1.116
0.912
19.44
19.88
14.38
18.89
13.89
10.35
19.26
19.46
14.10
18.30
13.55
10.14
18.97
18.89
13.75
17.70
13.22
9.88
19.28
19.73
14.99
18.77
14.38
11.60
18.69
18.61
14.13
17.50
13.50
10.86
18.68
18.60
14.10
17.47
13.47
10.84
19.26
19.71
15.30
18.74
14.60
12.12
18.64
18.56
14.36
17.46
13.67
11.32
18.64
18.54
14.33
17.44
13.64
11.27
69
Spans
NL
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
100
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
250
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
500
2
2
2
2
2
2
3
3
3
3
3
3
4
4
4
4
4
4
2
2
2
2
2
2
3
3
3
3
3
3
4
4
4
4
4
4
2
2
2
2
2
2
3
3
3
3
3
3
4
4
4
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
2
3
3
4
4
4
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
3
3
4
3
4
5
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
2.73
3.93
2.95
5.13
3.85
3.08
g Me +
gM
Eq. 4.1 Prog
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.819
1.255
0.941
1.691
1.269
1.015
0.758
1.148
0.891
1.541
1.188
0.986
0.758
1.148
0.891
1.542
1.189
0.987
0.758
1.148
0.891
1.542
1.189
0.987
0.721
1.100
0.838
1.481
1.128
0.918
0.721
1.101
0.838
1.483
1.130
0.919
0.721
1.101
0.838
1.483
1.130
0.919
0.705
1.079
0.813
1.453
1.101
0.886
0.705
1.079
0.813
1.455
1.102
0.887
0.705
1.079
0.813
1.455
1.102
0.887
g Mi +
g Me
g Mi
Diff. (%)
Prog
Diff. (%)
Prog
Diff. (%)
Prog
Diff. (%)
8.11
9.27
5.66
9.77
6.79
2.92
8.11
9.27
5.60
9.68
6.72
2.84
8.11
9.27
5.60
9.68
6.72
2.84
13.66
14.01
12.34
14.20
12.42
10.56
13.63
13.94
12.28
14.04
12.29
10.45
13.63
13.94
12.28
14.03
12.29
10.44
16.31
16.29
15.77
16.39
15.23
14.58
16.26
16.22
15.68
16.23
15.09
14.44
16.26
16.22
15.68
16.22
15.08
14.44
0.694
1.054
0.823
1.416
1.106
0.926
0.693
1.052
0.823
1.415
1.105
0.926
0.693
1.052
0.823
1.415
1.105
0.926
0.692
1.051
0.807
1.410
1.084
0.897
0.691
1.051
0.807
1.410
1.084
0.895
0.691
1.051
0.807
1.410
1.084
0.895
0.692
1.051
0.800
1.408
1.074
0.880
0.691
1.051
0.800
1.408
1.074
0.880
0.691
1.051
0.800
1.408
1.074
0.880
18.10
19.05
14.33
19.42
14.73
9.59
18.29
19.23
14.40
19.56
14.84
9.63
18.29
19.23
14.40
19.56
14.84
9.63
18.46
19.35
16.56
19.97
17.07
13.08
18.53
19.38
16.54
19.95
17.05
13.36
18.53
19.38
16.54
19.95
17.05
13.36
18.51
19.39
17.65
20.12
18.11
15.39
18.56
19.41
17.62
20.08
18.09
15.29
18.56
19.40
17.62
20.08
18.09
15.29
0.729
1.118
0.860
1.524
1.159
0.950
0.716
1.113
0.858
1.509
1.152
0.948
0.716
1.113
0.858
1.509
1.152
0.948
0.727
1.103
0.846
1.496
1.133
0.925
0.729
1.107
0.848
1.502
1.137
0.926
0.727
1.102
0.844
1.489
1.129
0.921
0.717
1.089
0.831
1.478
1.116
0.905
0.720
1.094
0.834
1.485
1.120
0.907
0.717
1.088
0.828
1.471
1.111
0.899
12.35
12.17
9.45
10.95
9.50
6.83
14.49
12.71
9.73
12.11
10.08
7.05
14.49
12.71
9.73
12.11
10.08
7.05
12.68
13.77
11.20
13.09
11.95
9.72
12.37
13.32
10.93
12.64
11.59
9.58
12.74
13.88
11.54
13.59
12.33
10.23
14.25
15.19
13.23
14.43
13.69
12.15
13.86
14.65
12.85
13.90
13.23
11.90
14.37
15.34
13.71
14.98
14.16
12.84
0.676
1.022
0.809
1.384
1.091
0.907
0.690
1.010
0.789
1.371
1.069
0.883
0.690
1.010
0.789
1.371
1.069
0.883
0.679
1.027
0.807
1.389
1.088
0.899
0.681
1.031
0.811
1.397
1.093
0.903
0.679
1.033
0.806
1.394
1.085
0.895
0.680
1.028
0.804
1.390
1.084
0.892
0.681
1.032
0.808
1.398
1.091
0.896
0.682
1.036
0.805
1.397
1.083
0.889
21.28
22.72
16.36
22.20
16.31
11.91
18.80
24.17
19.26
23.40
18.70
14.89
18.80
24.17
19.26
23.40
18.70
14.89
20.62
22.14
16.54
21.80
16.65
12.92
20.39
21.66
16.04
21.06
16.04
12.43
20.65
21.45
16.69
21.31
16.92
13.40
20.58
22.06
17.08
21.72
17.01
13.83
20.32
21.54
16.49
20.95
16.32
13.24
20.21
21.12
16.94
21.06
17.10
14.10
70
5.1 Introduction
Bridge officials are required to inspect all bridges periodically (biannually) to detect any
deficiencies or deterioration. The task of bridge inspection is especially difficult to perform
in box girder bridges. Almost half the inspection task is performed from inside the box due
to its tunnel-like nature. Other types of bridges (I-girders) do not require such extra effort as
all inspections are exterior in nature. The interior of the box, which can be dangerous
because of high temperatures and poor ventilation, is reached through access hatches that are
usually provided in the bottom flange immediately before or after an expansion joint. These
locations are chosen because: 1) bending moments are small close to the expansion joint, and
2) the abutment/pier over which the expansion joint is located facilitates access; inspection
crews can climb up the abutment or use simple tools such as a ladder to reach the access
hole. Since the spans covered by box girders are often long and the girders are constructed as
continuous segments over three or more supports, the distance between access hatches
frequently exceeds the limit that rescue crews can reach in the event of an emergency. The
addition of new access holes to an exiting bridge should satisfy certain practical constraints
(discussed shortly) and should not adversely affect the structural behavior of the bridge, i.e. it
should not impair serviceability nor decrease the ultimate strength and fatigue life. If the
access hole is placed in a critical location the bridge may need to be strengthened, which is
generally costly. Alternatively, access holes could be placed in low stress regions where
strengthening is not necessary thereby significantly reducing the cost of rehabilitation.
Holes in the bottom flange can significantly reduce the flexural strength of the cross section
and the entire structure, especially if they are placed in the vicinity of continuous supports
where high negative moments develop or in high positive moment regions. Away from high
moment regions, normal stresses due to flexure are small and adding a hole in the bottom
71
flange may be possible without strengthening. Web access holes can greatly reduce the shear
resistance of the web, which is made of relatively thin steel plates. Therefore, this alternative
should only be considered around midspan where low shear forces and torsional moments
exist. Since concrete deck dimensions are usually predetermined by traffic considerations
(road width) rather than by structural need, the effect of opening an access hole in the
concrete deck has the least impact on strength of all three alternatives.
Deck access
Web access
Reinforced
concrete deck
Steel plates
5.2.2 Feasibility
Access holes that will interfere with structural elements such as stiffeners and cross frames
should be avoided. Longitudinal bottom flange stiffeners are provided to stiffen the plate and
prevent buckling in the negative moment region over supports. These stiffeners are
discontinued in regions where the negative moment drops to zero. Narrow boxes are
provided with one stiffener in the middle of the plate, which creates an obstacle to adding
access holes in the bottom flange near a support. Boxes with two bottom flange stiffeners can
be provided with an access hole between the stiffeners if the distance in between the
stiffeners is sufficient. Web access holes can be added in the space between web stiffeners.
Top flange bracing may block access through reinforced concrete deck hatches. Since,
braces are provided to stabilize the steel cross-section before casting the concrete deck they
can be taken out if needed since they are not essential for the behavior of the steel/concrete
composite closed cross section. Cross frame locations should be avoided since these
structural components are important for the global behavior of the bridge.
5.2.3 Accessibility
Accessibility of access holes is an important issue since they are added for use in the event of
an emergency. Holes in the concrete deck are easily accessible from the roadway and do not
72
need special equipment. Bottom flange and web alternatives that are away from the supports
require special equipment such a snooper (truck with an arm that reaches over the side and
below the bridge) or a bucket (which reaches from below). In the vicinity of the supports,
piers and abutments facilitate access to the holes. In the case of multi cell bridges, web
openings in inner cells will be hard to reach if they are located away from the piers.
5.2.4 Water Leakage
Any modifications in the bridge that would increase the potential for corrosion should be
avoided. Access holes in bottom flanges and webs (except for outer webs where water may
run down the sides) do not increase the chances of water getting into the steel box. The
concrete deck alternative is the most critical of all three alternatives since imperfect hatch
doors may cause water leakage into the box.
5.2.5 Impact on Traffic
Access holes that are accessed from the topside of the bridge (concrete deck, or web and
bottom flange holes accessed with the help of a snooper) will impede traffic flow and need
special arrangements to use the access hatch. No such arrangements are needed for bottom
flange and web openings that will be accessed from beneath the bridge using a bucket or
directly using the piers.
5.2.6 Unauthorized Access
Unauthorized people can easily reach access holes in the vicinity of the supports and in the
concrete deck. Away from the supports, web or bottom flange access is not possible without
the use of special equipment. In all cases, precautions should be taken to prevent
unauthorized people from getting into the cells. Figure 1.2 shows the belongings of an
individual who made the bridge cell his home.
Based on the previous discussions it is clear that the concrete deck option is the least
attractive alternative because of the potential for water leakage to the inside of the girder and
the possibility of interference with traffic flow when in use. The most appealing option is
bottom flange access, which is practical and, as will be discussed later on in the report, does
not have an impact on bridge strength if the location is chosen appropriately. The web option
has a greater number of limitations, which makes it a last resort if the bottom flange option is
not possible. Table 5.1 summarizes the advantages and disadvantages of each of the
alternatives.
73
Bottom Flange
Web
Concrete Deck
Strength
Reduces flexural
strength.
Feasibility
Stiffeners should be
avoided. May be an
obstacle if one stiffener
is used.
Stiffeners must be
avoided.
No stiffener limitations,
but construction bracing
may be an obstacle.
Easy
Easy
Access between
supports
Water leakage
Not a concern.
Impact on traffic
No effect.
No effect.
Unlawful Access
Possible at supports.
Possible at supports.
whose capacity would still exceed the demand even with the new hole. The following
sections describe the details of the approaches followed in this research.
The previous approaches are applied to the results obtained from the analysis of an idealized
bridge model for the sake of demonstration. The idealized bridge is a three span continuous
box girder. The length of each span is 75 ft. Each approach yields regions of low stresses in
accordance with the assumptions made in the scheme. Figure 5.2-a shows the regions that
satisfy the normal stress thresholds discussed in Approach I (without considering fatigue
limitations). As expected, minimally stressed regions exist around the points of contraflexure
and close to exterior supports. Approach II results in the regions seen in Fig. 5.2-b, which
are located around mid-spans where low shear forces and torsional moments exist. The
fatigue stress distribution for the idealized bridge is given in Fig. 5.3-a. The empty regions
represent areas where stresses are compressive and are therefore not affected by fatigue. It
turns out that for this particular bridge fatigue stresses are not critical anywhere along the
bridge and therefore do not limit where access holes can be placed (Fig. 5.3-b). The regions
suitable for access hole placement, i.e. that satisfy both strength and fatigue criteria, are
obtained by imposing the regions in Figures 5.2-a and 5.3-b and are shown in Figure 5.4.
75
1st Span
2nd Span
3rd Span
60
<45% region
<33% region
<20% region
50
40
30
20
10
0
0
75
150
Longitudinal Position on Girder (ft)
225
(a)
60
<45% region
<33% region
<20% region
50
40
30
20
10
0
0
75
150
225
(b)
Figure 5.2: Low stress regions for idealized bridge (a - Approach I, b - Approach II).
76
1st Span
2nd Span
60
3 rd Span
AASHTO limit
50
40
30
20
10
0
75
150
Longitudinal Position on Girder (ft)
225
(a)
1st Span
2nd Span
75
3rd Span
150
225
(b)
Figure 5.3: (a) Distribution of fatigue stresses. Empty regions are areas in compression
and therefore not affected by fatigue. (b) Regions not critical for fatigue considerations
(entire bridge).
77
1st Span
2nd Span
75
150
Longitudinal Position on Girder (ft)
3 rd Span
225
Figure 5.4: Regions that satisfy both normal stress criteria (<33% regions in Figure 5.2a and fatigue stress criteria in Fig. 5.3-b). Shaded regions are suitable for access hole
placement.
The chosen bridges are studied using the above-described approaches to investigate whether
low stress regions as defined earlier can be identified. The bridges are located in District IV
of the Florida Department of Transportation and vary in dimensions covering a range of
curvatures and cross sections. All bridge segments are studied. Based on the analyses
conducted in this study the following results are obtained. Details for each individual bridge
can be found in Okeil et al (2000).
Approach I: For spans that are continuous at both ends, the location of minimally stressed
points is at a distance ranging from 20% to 42% of the span length measured from the
continuous support. Corresponding values for first/last spans are 20% to 54% measured from
the continuous support. The length of the minimally stressed region varies substantially. On
average, the length of the region is 5.6% of the span length, L. Spans that are continuous on
both sides have an average length of 6.0%L, and spans that are continuous from one side
have an average region length equal to 4.7%L. Examination of the results showed that there
was no clear correlation between horizontal curvature and location or size of the minimally
stressed regions. The lack of a definite trend is attributed to 1) the limited number of case
studies considered; 2) the complexity of the structural problem; and 3) the fact that the
bridges were designed by different design firms. Different design practices can lead to some
variations in design, which can substantially affect the position and size of the minimally
stressed regions.
Approach II: Locations with minimal shear stress exist mainly around the middle of each
span where low shear forces and torsional moments act on the girder. It is observed that for
first/last spans appropriate regions are at an average distance of 37%L measured from the end
support, and for spans continuous on both ends, the average distance is 50%L. The average
region length for first/last spans and spans continuous on both sides is found to be 40%L and
78
79
Hole location
Inner supports
(tension)
steel
Fy
(compression)
concrete
steel
ft
concre
te
0.005
'
(tension)
fc
Fy
(compression)
80
Figure 5.7: Bottom view of bridge model showing hole location in a minimally stressed
region.
81
Current codes pertaining to analysis and design of curved composite girders are mostly based
on experimental and analytical research conducted over 30 years ago as part of project CURT
(Consortium of University Research Teams) funded by the Federal Highway Administration
(FHWA). A new Curved Steel Bridge Research (CSBR) project is currently being conducted
under the auspices of the FHWA. Although the CSBR project is expected to provide much
needed information on behavior, analysis and design of curved composite bridges, it focuses
more on I-girders than on box girders. The overall objective of the research reported herein is
to provide information that complements existing data and that will be useful for formulating
comprehensive design guidelines for composite curved box girders. Specific objectives
include: 1) investigate and quantify the effect of nonuniform torsion on the behavior and
design of existing curved box girders; 2) investigate existing distribution factors for curved
box girder bridges; and 3) provide information that is helpful in identifying suitable locations
for placement of access hatches in the steel box.
A computer program for simulating the behavior of curved box girders was developed. The
program is graphically interactive and features a general purpose beam-column element that
can account for the effect of warping. The developed program is used to conduct a detailed
investigation of warping related stresses in eighteen existing box girder bridges chosen from
the Florida Department of Transportation inventory. The bridges are carefully selected to
cover a wide range of design parameters including horizontal curvature, cross sectional
properties, and number of spans. They were designed by different firms and were constructed
at different times and are considered to be representative of current design practice. Forces
are evaluated from analyses that account for the construction sequence and the effect of
warping. Loading is considered following the 1998 AASHTO-LRFD provisions.
By considering differences between stresses obtained taking into account warping and those
calculated by ignoring warping, it is shown that warping has little effect on both shear and
normal stresses in the limited sample of bridges considered. These results should not be
construed to imply that warping is not important. Rather, this work points out that there could
be a large subset of bridges where the warping effect is small enough to be ignored in
structural calculations. This is particularly useful to designers because warping calculations
are complicated and time consuming.
Another study was undertaken to investigate load distribution factors promoted by current
specifications. Single girder and detailed grillage models were created for a variety of
bridges and analyzed. The parameters investigated are the number of girders, roadway width
expressed by number of lanes, girder spacing, span length, and radius of horizontal curvature.
For the limited pool of bridges studied, it was observed that errors of up to 25% can occur
when current DF expressions are used.
82
Access hatches (holes) in curved box girder bridges are usually provided in the bottom flange
immediately before or after an expansion joint. If additional access hatches are required after
the bridge is built, they must be placed in such a way that 1) they satisfy important practical
constraints such as feasibility, accessibility, water leakage, traffic impact, and unauthorized
access; 2) they do not adversely affect the structural behavior of the bridge, i.e. their
installation should not impair serviceability nor decrease ultimate strength or fatigue life.
Both issues are discussed and approaches that are suitable for identifying appropriate
locations for access hole placement are identified. Access hatches can be installed without
additional strengthening in low stress regions in the bottom steel flange. Low stress regions
can be found using the computer program developed in this research.
83
7 REFERENCES
84
Johnson, S.B. and A.H. Mattock (1967), Lateral Distribution of Load in Composite Box
Girder Bridges, Highway Research Record No. 167, Bridges and Structures.
McGuire W., Gallagher R.H., and R.D. Ziemian (2000), Matrix Structural Analysis, 2nd
Edition, John Wiley & Sons, Inc.
Nakai, H., and C.P. Heins (1977), Analysis Criteria for Curved Bridges, Journal of
Structural Engineering, ASCE, Vol. 103, No. 7.
Nakai, H., and C.H. Yoo (1988), Analysis and Design of Curved Steel Bridges, McGraw Hill,
New York, N.Y.
Okeil, A.M., S. El-Tawil, and M. Chaphalkar (2000), New Access Hatches in Existing
Curved Box Girder Bridges: A study based on practical and structural considerations,
Research Report, Department of Civil and Environmental Engineering, University of
Central Florida, Orlando, Florida.
Okeil, A.M., and S. El-Tawil (Feb. 2002), Considerations for Opening New Access Holes in
Curved Box Girders, Practice Periodical of Structural Design and Construction,
ASCE, Vol. 7, No. 1, pp 26-36.
Oleinik, J.C. and Heins, C.P. (Oct. 1975), Diaphragms for Curved Box Beam Bridges,
Journal of Structural Engineering, ASCE, Vol. 101, No. ST10, pp 2161-2178.
Sennah, K.M., and J.B. Kennedy (Jun. 2001), State-of-the-Art in Design of Curved BoxGirder Bridges, Journal of Bridge Engineering, ASCE, Vol. 6, No. 3, pp 159-167.
Sennah, K.M., and J.B. Kennedy (Mar. 2002), Literature Review in Analysis of Box-Girder
Bridges, Journal of Bridge Engineering, ASCE, Vol. 7, No. 2, pp 134-143.
Trukstra, C.J., and A.R. Fam (Mar. 1978), Behavior Study of Curved Box Bridges, Journal
of the Structural Division, ASCE, Vol. 104, ST3, pp 453-462.
Waldron, P. (1988), The Significance of Warping Torsion in the Design of Straight
Concrete Box-girder Bridges, Canadian Journal of Civil Engineering, Vol. 15, pp 879889.
Zureick, A. Naqib, R. and Yadlosky, J. M. (1994), Curved Steel Bridge Research Project;
Interim Report I: Synthesis, Publication No. FHWA-RD-93-129, Office of Engineering
and High Way Operation R&D, Federal Highway Administration, 6300 Georgetwon
Pike, McLean, VA 22101.
85
Zureick, A., Linzell, D., Leon, R. T., and Burell, J. (2000), Curved Steel I-Girder Bridges:
Experimental and Analytical Studies, Engineering Structures, Vol. 22, pp. 180-190.
86
For a thin-walled open section, the shear center, S , and the centroid, O , of the cross section
may not coincide. The first step of finding the geometric properties is to find the coordinates
of the shear center using the following relationships:
xS =
y dA
y dA
(8.1a)
xS =
y dA
y dA
(8.1b)
where o is an initially assumed warping function obtained with the centroid of the cross
section as an origin using the following expression
( s) = 1 ( s ) + 1 (0)
(8.2)
The first term of the previous expression is an integral of the perpendicular distance, , from
the point under consideration (centroid in the initial attempt) to the tangent to the point on the
cross section.
s
1 ( s ) = ds
(8.3)
1 (0) =
1
1 ( s )dA
A
(8.4)
After determining the shear center, the warping function is computed, considering the shear
center this time, following Eq. 8.2. The warping constant, I , is then obtained using the
following integral
I = (s ) (s )dA
(8.5)
87
In the case of a closed cross section as the one shown in Fig. 8.1, Eq. 8.2 becomes
s
( s) =
ds + 1 (0)
t
(8.6)
which may be viewed as two warping functions superimposed together (see Fig. 8.2). Once
( s) is determined, Eq. 8.5 is used to obtain I .
The sectoral area for a closed section is obtained using the following expression
S ( s) = S o ( s)
ds
1
S o ( s)
t
(ds t )
(8.7)
In which S o ( s) is the sectoral area for the same cross section assuming a slit that converts it
into an open cross section. Though other tools are available for determining S x and S y , an
approach similar to the one described for S was followed.
The torsional constant used in all analyses was determined based on the following
expression:
J=
4 Ac2
ds
t
(8.8)
which estimates torsional rigidity for closed cross sections and ignores the negligible
contributions of elements not enclosing an are such as the deck overhangs.
88
t
S (xs, ys)
ds
89
90